Comparison between two-dimensional and three-dimensional imaging utilizing the Grummons frontal analysis

INTRODUCTION

In their private practice, orthodontists primarily focus on the sagittal dimension when assessing patients. Still, the frontal dimension, as patients perceive it, deserves greater attention not only for correcting malocclusion but also for achieving functional balance and harmony between facial and smile proportions.^[1-3]

At present, two imaging modalities are used to evaluate the craniofacial anteroposterior dimensions: Conventional posteroanterior (PA) cephalograms and cone-beam computed tomography (CBCT) coronal plane. Previous studies comparing conventional PA cephalograms and three-dimensional (3D) volumes have consistently shown that CBCT imaging excels in assessing the frontal dimension.^[4-7] To integrate CBCT effectively into daily orthodontic practice, dental imaging software has been developed to generate cephalograms based on CBCT volumes. This is in response to the lack of practical, reliable 3D analyses, which hinder direct application to 3D volumetric images.^[8-10]

To date, no study has yet compared the outcomes between CBCT-constructed PA cephalograms and tomographic slices. Thus, our study aimed to compare the frontal dimensions between CBCT-synthesized PA cephalograms and multiplanar reconstruction (MPR) slices using Grummons’ frontal analysis to determine the extent of agreement between the two imaging techniques while adjusting for confounding variables encountered in 2D projections: head orientation and magnification. This would assert whether CBCT-reconstructed PA is reliable for the assessment of the frontal dimension, given the lack of adequate 3D analyses, precise definition of 3D landmarks, practicality, time saving, and radiation safety considerations.

MATERIAL AND METHODS

Study design

This retrospective cross-sectional study was approved by the Lebanese University Scientific Committee in Beirut, Lebanon. The sample size was determined based on a significance level of 0.05 and a power of 80%, indicating that 34 patients were required. Pre-treatment CBCT scans of 39 patients (22 females and 17 males) were retrieved from the Department of Orthodontics and Dentofacial Orthopedics at the Lebanese University. The study inclusion criteria were patients aged range between 8 and 13 years, without any craniofacial syndrome neither periodontal diseases nor missing permanent upper first molars, or who had undergone prior orthodontic treatment.

Low-dose large field of view CBCT scans were captured using the iCat machine with consistent settings: 120kVp, 5 mA, field of view: 13 × 17 mm, voxel spacing: 0.4 mm, and scan time: 20 s. Scans were acquired with patients in maximum intercuspation, aligning the Frankfurt and Midsagittal planes parallel and perpendicular to the ground, respectively. Data were reconstructed with a 0.4 mm slice thickness, saved as digital imaging and communications in medicine files, and imported into Viewbox^® version 4.0.1.6, Athens, Greece.^[11]

A total of 26 skeletal landmarks were identified on CBCT scans using MPR slices and 3D volume reconstructions [Figures 1-4]. The Frankfurt plane and Midsagittal reference planes were automatically generated by plotting three landmarks for each [Table 1]. Subsequently, scans were positioned to ensure that the constructed skeletal Frankfort plane was parallel to a horizontal line in both sagittal and frontal views. This was done to eliminate head positioning errors when generating two-dimensional (2D) constructed PA cephalograms. The generation method employed the RayCast technique, which displays several MPR slices combined to produce an image representing the specific patient’s volume needed to generate constructed 2D projections. The resultant images are free of magnification or distortion. However, the overlapping of skeletal structures was not resolved.^[12] Twenty-one landmarks were mapped exclusively to 2D projections according to the definitions in [Table 1 and Figure 5].

Close

Figure 1:

(a) Construction of the midsagittal plane connecting crista galli, sella turcica through anterior nasal spine and extended several millimeters inferior to the chin. (b) Construction of the Frankfort plane; plane connecting the right and left Porion and the midpoint between right and left Orbitale.

Export to PPT

Close

Figure 2:

(a) Localization of point (Me); on the sagittal view of the multiplanar reconstruction (MPR), the frontal and axial planes intersected at the most inferior point of the symphysis. (b) Then, point (Me) was digitized on the frontal view on the MPR. (c) Localization of point Condylion (Co); identified by scrolling supero-inferiorly on the coronal view of the MPR until localizing the most superior point of the condyle. (d) Localization of point Antegonion (AG), on the lateral view of the 3D volume rendering, the frontal and axial planes intersected at the deepest aspect of the antegonial notch. (e) Then, points (AgR and AgL) were digitized on the frontal view of the MPR.

Export to PPT

Close

Figure 3:

(a) Localization of the point anterior nasal spine (ANS). (b) By scrolling through the sagittal view of the MPR, the frontal and axial planes intersected at the tip of ANS. (c) Then point (ANS) was digitized on the axial view of the multiplanar reconstruction.

Export to PPT

Close

Figure 4:

(a) Localization of Juguale points bilaterally (JR and JL). (b) On the frontal view of the 3D volume rendering, the sagittal and axial planes intersected at the deepest aspect of the zygomatic process of the maxilla. Then, points (JR and JL) were digitized on the frontal view of the multiplanar reconstruction.

Export to PPT

Close

Figure 5:

On cone-beam computed tomography reconstructed posterioanterior cephalograms: (a) generation of reference planes and (b) identification on two-dimensional landmarks.

Export to PPT

Table 1: Cephalometric landmarks and reference planes located on two-dimensional and three-dimensional analyses.

Landmarks	Definition
Crista galli	The most superior point of Crista galli[13]
Anterior nasal spine	Most anterior point of the anterior nasal spine[14]
Menton	The most inferior point of the symphysis of the mandible[14]
Upper dental midline	The most marginal point at the incisal level between the upper centrals[14]
Lower dental midline	The most marginal point at the incisal level between the lower centrals[15]
Sella turcica	The center of Sella Turcica[14](on 3D only)
Zygomatico-frontal suture	Medial junction of zygomatic bone with the frontal bone[4]
Zygomatic arch	The outermost point of the zygomatic arch[14]
Juguale process	The intersection of the outline of the maxillary tuberosity and the zygomatic buttress[5]
Nasal cavity	The most lateral point of the bony nasal cavity[10]
Condylion	The most superior point of the condyle[11]
Antegonion	The deepest point on the curvature of the antegonial notch[5]
Mesiobuccal cusp of upper maxillary first molar	The tip of the mesiobuccal cusp of the maxillary first permanent molar[3,5]
Porion	The highest point of the external auditory meatus[14](on 3D only)
Orbitale	The lowest point in the inferior margin of the orbit[14](on 3D only)
Midsagittal reference plane on 2D	Line connecting crista galli and anterior nasal spine[3][Figure 5]
Midsagittal reference plane on 3D	Plane connecting crista galli, anterior nasal spine, and sella turcica[12][Figure 1]
Frankfort plane on 3D	Plane connecting the right and left porion and the midpoint between right and left orbitale[14][Figure 1]

All 3D scans and their corresponding 2D constructed PA cephalograms were analyzed for each patient using preprogrammed 3D and 2D simplified forms of Grummons analysis, respectively [1-3] . These forms were integrated into the Viewbox software. Finally, all 2D and 3D linear, angular, and ratio measurements for each patient were comprehensively analyzed.

RESULTS

High intra-class correlation coefficients indicated strong interobserver reliability across all parameters. The Bland– Altman test results showed no correlation for the following variables: Condylion-Antegonion-Menton Right (CoAgMeR), Condylion-Antegonion-Menton Left (CoAgMeL), Condylion Right (CoR), Condylion Left (CoL), Juguale Right (JR), Juguale Left (JL), Condylion - Antegonion Right (CoAgR), Condylion - Antegonion Left (CoAgL), Condylion Menton Right (CoMeR), Condylion - Menton Left (CoMeL), Menton - Antegonion Right (MeAgR), Menton - Antegonion Left (MeAgL), Upper Facial Ratio (UFRa), Lower Facial Ratio (LFRa), Maxillary Ration (MxRa) . As shown in [Tables 2 and 3], significant differences were found in linear measurements involving points (Co), (J), and especially (Me) (P < 0.05). Inter-point distances between (Co), (Ag), and (Me) were also significant, with the largest discrepancy between (Co) and (Me) bilaterally (P < 0.05). Among angular parameters, only CoAgMe differed significantly (P < 0.05). Ratios varied significantly between imaging techniques, except for the total maxillary ratio (P < 0.05). Furthermore, Cohen’s d-test was applied to confirm clinical significance. Although a strong effect size was observed in most variables, there was a weak to medium effect size in seven instances where statistical significance did not translate into clinical relevance.

DISCUSSION

Inaccurate landmark identification remains a major source of error in cephalometric studies.^[13-16] To minimize this, strict control of head orientation and magnification was applied during the generation of CBCT-based PA cephalograms, ensuring consistent results between the two imaging techniques. Nevertheless, structural superimposition continues to pose a limitation in 2D radiographs.^{[11,15,17,18,19]}

The observations of Sfogliano et al., Shokri et al., and Oz et al. align with the present findings regarding the Condylion and Menton landmarks.^[10,20,21] These authors highlighted the potential for error when identifying landmarks located on curved surfaces in 2D radiographs, supporting the advantage of MPR slices for detecting the most convex points of the condyles and chin. Conversely, in 2D images, overlapping anatomical structures near the condyles often hinder the accurate identification of Condylion.^[10,20,21]

Regarding the antegonion point, our results showed it to be the most difficult to locate on MPR slices due to challenges in determining the deepest concavity along the mandibular lower border. Similar observations were reported by Tai et al. and Cheung et al., who noted that in 2D projections, the superimposition of the posterior and anterior mandibular borders facilitated the apparent localization of this landmark.^[22,23] In the present study, antegonion was marked by superimposing MPR slices onto the volume rendering, which may explain the absence of a statistically significant difference between the two imaging modalities. This also accounts for the variations in linear and angular measurements related to mandibular symmetry assessment.^[5] Unlike 2D projections, where Condylion, antegonion, and menton appear on a single tomographic plane, 3D analysis considers them across different planes. The lack of an anteroposterior component in 2D projections consequently leads to greater discrepancies in angular measurements.^[13]

In addition, a statistically significant difference was observed between the two imaging methods for linear measurements involving the Juguale point. These findings correspond with those of Shokri et al., Tai et al., and Cheung et al.^[20,22,23] Since Juguale is defined as the intersection of the maxillary tuberosity and the zygomatic arch, this definition becomes unsuitable in 3D, where such intersections are not spatially defined. Therefore, landmarks traditionally defined in two dimensions do not fully represent their true 3D positions, complicating their identification in volumetric analyses.^[13]

CONCLUSION

In summary, we can state that MPR slices and CBCT constructed PA were not comparable. Statistically significant differences in points Juguale, Condylion, and Menton suggest errors in diagnosing transverse discrepancies and in assessing mandibular asymmetry between the two techniques. Thus, new 3D analyses must be developed to accurately reflect the 3D reality and compared to gold standard measurements, which are those made on dry skulls. It has been proposed that a shift from landmarks, distances, and angles to volumes, areas, and surfaces would provide a more accurate diagnosis.